Special equipment state recognition method and system based on electrical parameter pattern matching

Abstract

The application discloses a special equipment state recognition method and system based on electrical parameter mode matching, and relates to the technical field of data processing.The method comprises the following steps: verifying the state sensitivity of multiple monitoring indexes of a power distribution cabinet to obtain a dynamic grading sensitive index set; generating an elevator state mode library; calling multiple real-time operation parameters; triggering hierarchical equipment state mapping to output a hierarchical state confidence vector; performing state transition prediction to output a state transition probability matrix; fusing the hierarchical state confidence vector and the state transition probability matrix to output a real-time fault probability distribution; and generating a multi-stage early warning signal according to the real-time fault probability distribution to trigger a step-by-step safety protection response.The technical problem that the state recognition of special equipment in the prior art relies on sensors is solved, the accurate recognition of the running state of an elevator based on electrical parameters is realized, the dependence on external sensors is reduced, and the technical effects of improving the real-time performance and reliability of state monitoring are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a special equipment state recognition method and system based on electrical parameter pattern matching. BACKGROUND

[0002] In the running process of special equipment, safety state monitoring and fault identification are important links to ensure stable operation of the equipment and safety of personnel. In the prior art, the running state recognition of an elevator mainly relies on multiple types of sensors installed at key parts of the equipment, which analyzes physical quantities such as vibration, temperature, and displacement. However, this method has the following disadvantages: on the one hand, the installation and maintenance cost of the sensors is high, and the sensors are easily affected by factors such as installation position, environmental interference, and performance degradation of the sensors themselves, which makes it difficult to ensure the stability and accuracy of the monitoring data; on the other hand, the change law of electrical parameters under different running states has not been fully utilized, and the existing scheme lacks pattern recognition capability based on electrical parameters, making it difficult to continuously achieve accurate state recognition and trend prediction in the case of limited or failed sensors. SUMMARY

[0003] The present application provides a special equipment state recognition method and system based on electrical parameter pattern matching, which solves the technical problem of relying on sensors for special equipment state recognition in the prior art.

[0004] In a first aspect, the present application provides a special equipment state recognition method based on electrical parameter pattern matching, which comprises:

[0005] Verifying the state sensitivity of multiple monitoring indicators of a power distribution cabinet to obtain a dynamic hierarchical sensitive indicator set; pre-constructing a multi-level state pattern sub-library offline according to the dynamic hierarchical sensitive indicator set, and generating an elevator state pattern library through hierarchical association integration; according to the dynamic hierarchical sensitive indicator set, gradually calling multi-level real-time running parameters from the power distribution cabinet in order of sensitivity priority; dynamically loading the multi-level real-time running parameters into the elevator state pattern library, triggering hierarchical equipment state mapping, and outputting a hierarchical state confidence vector; performing state transition prediction on the hierarchical state confidence vector, and outputting a state transition probability matrix; fusing the hierarchical state confidence vector and the state transition probability matrix, and outputting a real-time fault probability distribution; generating a multi-level early warning signal according to the real-time fault probability distribution, and triggering a step-by-step safety protection response.

[0006] In a second aspect, the present application provides a special equipment state recognition system based on electrical parameter pattern matching, which comprises:

[0007] The sensitivity verification module verifies the sensitivity of the multiple monitoring indexes of the power distribution cabinet to the state of the special equipment, and obtains a dynamic hierarchical sensitive index set; the mode library construction module pre-constructs a multi-level state mode sub-library according to the dynamic hierarchical sensitive index set, and integrates the multi-level state mode sub-library through hierarchical association to generate an elevator state mode library; the parameter calling module calls multi-level real-time operation parameters from the power distribution cabinet according to the dynamic hierarchical sensitive index set, in a hierarchical manner according to the sensitivity priority; the state mapping module dynamically loads the multi-level real-time operation parameters into the elevator state mode library, triggers hierarchical equipment state mapping, and outputs a hierarchical state confidence vector; the state transition prediction module performs state transition prediction on the hierarchical state confidence vector, and outputs a state transition probability matrix; the fault probability output module fuses the hierarchical state confidence vector and the state transition probability matrix, and outputs a real-time fault probability distribution; and the early warning response module generates a multi-level early warning signal according to the real-time fault probability distribution, and triggers a hierarchical safety protection response.

[0008] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] First, the sensitivity of the multiple monitoring indexes of the power distribution cabinet to the state of the special equipment is verified, and a dynamic hierarchical sensitive index set is obtained. Then, a multi-level state mode sub-library is pre-constructed according to the dynamic hierarchical sensitive index set, and the multi-level state mode sub-library is integrated through hierarchical association to generate an elevator state mode library; at the same time, multi-level real-time operation parameters are called from the power distribution cabinet according to the dynamic hierarchical sensitive index set, in a hierarchical manner according to the sensitivity priority. Further, the multi-level real-time operation parameters are dynamically loaded into the elevator state mode library, hierarchical equipment state mapping is triggered, and a hierarchical state confidence vector is output. Next, state transition prediction is performed on the hierarchical state confidence vector, and a state transition probability matrix is output. Then, the hierarchical state confidence vector and the state transition probability matrix are fused, and a real-time fault probability distribution is output. Finally, a multi-level early warning signal is generated according to the real-time fault probability distribution, and a hierarchical safety protection response is triggered. The technical problem that the state recognition of the special equipment in the prior art relies on sensors is solved, the precise recognition of the running state of the elevator based on electrical parameters is achieved, the dependence on external sensors is reduced, and the technical effects of improving the real-time performance and reliability of state monitoring are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0011] Figure 1A flowchart of a special equipment state recognition method based on electrical parameter pattern matching is provided for the embodiments of the present application.

[0012] Figure 2 A structural diagram of a special equipment state recognition system based on electrical parameter pattern matching is provided for the embodiments of the present application.

[0013] Legend: sensitivity verification module 11, pattern library construction module 12, parameter calling module 13, state mapping module 14, state transition prediction module 15, fault probability output module 16, and early warning response module 17. DETAILED DESCRIPTION

[0014] The present application provides a special equipment state recognition method and system based on electrical parameter pattern matching, which solves the technical problem that the state recognition of special equipment in the prior art relies on sensors.

[0015] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0016] It should be noted that the terms "comprise" and "have" are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server comprising a series of steps or units need not be limited to only those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to the process, method, product, or device.

[0017] Embodiment one, as shown in the present application provides a special equipment state recognition method based on electrical parameter pattern matching, wherein the method comprises: Figure 1

[0018] The sensitivity of the multiple monitoring indicators of the power distribution cabinet to the state of the special equipment is verified to obtain a dynamic hierarchical sensitive indicator set.

[0019] ​Based on the application environment characteristics and equipment model code of the power distribution cabinet, a device fingerprint is constructed, which is used as a retrieval condition to call full-dimensional operation log data of multiple devices of the same type. Then, the collected operation logs are aggregated according to the index type to form multiple state time series data sets corresponding to multiple monitoring indexes. For the multi-state time series data of each monitoring index, the data is divided using the device state label to construct a reference state record set and multiple fault state record sets. By calculating the KL divergence of the reference state and each fault state record set, the sensitivity scores of each monitoring index to different fault states are quantified. Combined with the recurrence frequency of the fault state, the sensitivity scores of each fault state are weighted and summarized to obtain the comprehensive operation sensitivity of each monitoring index. Finally, the monitoring indexes are dynamically classified according to the comprehensive operation sensitivity to form a dynamically classified sensitive index set.

[0020] Further, the sensitivity of multiple monitoring indexes of the power distribution cabinet to special equipment states is verified to obtain a dynamically classified sensitive index set, and the method comprises:

[0021] The application environment characteristics of the power distribution cabinet are retrieved, and the application environment characteristics and the power distribution cabinet model code are used as a device fingerprint coding retrieval condition to call full-dimensional operation log data of multiple devices. Based on the index type, the full-dimensional operation log data of the multiple devices is aggregated to obtain multiple multi-state time series data sets of multiple operation indexes. Based on the multiple multi-state time series data sets, the sensitivity of special equipment states is quantified, and multiple operation comprehensive sensitivities are output. According to the multiple operation comprehensive sensitivities, the multiple operation indexes are dynamically classified, and the dynamically classified sensitive index set is output.

[0022] Firstly, the application environment characteristics of the target power distribution cabinet are retrieved, which include but are not limited to installation location, power supply topology, load type, environmental temperature and humidity range, etc., and the power distribution cabinet model code is obtained; the above-mentioned application environment characteristics and model code are combined to generate a unique device fingerprint code, which is used as a retrieval condition to call the full-dimensional operation logs of multiple similar power distribution cabinets matching the device fingerprint from the historical operation database, the operation logs including time sequence records of each monitoring index under multiple operating states. Then, the full-dimensional operation logs of the multiple devices are aggregated according to the index type, and the monitoring data of the same type is classified into the corresponding operating index to form multiple state time sequence data sets corresponding to multiple operating indexes, wherein the multiple states include a reference state and several fault states, and each state corresponds to the monitoring record of one or more time periods. Next, based on the multiple multi-state time sequence data sets, state sensitivity quantitative analysis is performed on each operating index; specifically, taking the index distribution characteristics of the reference state as a reference, the distribution difference measurement value between the index in each fault state and the reference state is calculated respectively, and the measurement can use Kullback-Leibler divergence (KL divergence), Jensen-Shannon divergence (JSD) or other statistical distance algorithms; the difference measurement values of each fault state are weighted and summed according to the historical recurrence frequency of the fault state to obtain the operation comprehensive sensitivity of the operating index. Finally, the operation comprehensive sensitivity of all operating indexes is sorted in descending order, and is dynamically classified according to the preset classification threshold or quantile interval, and the indexes with higher sensitivity are classified into high sensitivity level, the indexes with medium sensitivity are classified into medium sensitivity level, and the indexes with lower sensitivity are classified into low sensitivity level; the dynamic classification sensitive index set containing index identifier, sensitivity score, sensitivity level and main associated fault type is output for subsequent state pattern construction and real-time state recognition.

[0023] Further, based on the multiple multi-state time sequence data sets, the state sensitivity of special equipment is quantified, and multiple operation comprehensive sensitivities are output, the method comprising:

[0024] Based on the device state label, the multiple multi-state time sequence data sets are decomposed to obtain multiple reference state record sets and multiple fault state record sets; taking the multiple reference state record sets as a reference, the KL divergence is used to quantify the multiple fault state record sets, and multiple fault state KL sensitivity scores are output; according to the recurrence frequency of M fault states, the multiple fault state KL sensitivity scores are weighted within the group, and the multiple operation comprehensive sensitivities are output.

[0025] Firstly, the multiple multi-state time series data sets are state-decomposed using the device state labels corresponding to each multi-state time series data set, data in a normal operating state is divided into a reference state record set, and data in different types of fault states is divided into multiple fault state record sets, each fault state record set corresponding to one fault type. Then, taking multiple reference state record sets as reference references, the distribution difference between each fault state record set and the corresponding reference state record set of each operating index is measured; specifically, the reference state record set and the fault state record set are respectively subjected to probability density estimation to obtain reference distribution and fault distribution, and the distribution difference between the two is calculated using Kullback-Leibler divergence (KL divergence) to obtain the KL sensitivity score of the operating index under the fault state. Next, according to the recurrence frequency of M fault states in the historical operation data, the corresponding weight coefficients are constructed, which are proportional to the fault recurrence frequency and normalized; the KL sensitivity score of each operating index under each fault state is weighted and summed according to the corresponding weight coefficient to obtain the operation comprehensive sensitivity of the operating index. Finally, the operation comprehensive sensitivity of all operating indexes is output as the input basis for the generation of the subsequent dynamic hierarchical sensitive index set.

[0026] According to the dynamic hierarchical sensitive index set, a multi-level state mode sub-library is pre-constructed offline, and an elevator state mode library is generated through hierarchical association and integration.

[0027] Based on the dynamic hierarchical sensitive index set, the sensitive indexes are decomposed according to the hierarchical relationship to obtain a multi-level sensitive index set. Taking the first-level sensitive index as a screening condition, a corresponding plurality of sample fault state time series data sets are retrieved from the device full-dimensional operation log, and combined with different fault level labels to form a plurality of multi-state sample data sets. The sample data is subjected to index fluctuation scale analysis, representative state feature templates are extracted, and a first-level state mode sub-library is constructed. Subsequently, according to the hierarchical order of the sensitive indexes, second-level and higher-level state mode sub-libraries are iteratively constructed, and a bidirectional mapping index between the multi-level sensitive indexes and the multi-level state mode sub-libraries is established. Finally, through hierarchical association and integration, a complete elevator state mode library is generated, providing a structured mode basis for subsequent real-time state matching and recognition.

[0028] Further, according to the dynamic hierarchical sensitive index set, a multi-level state mode sub-library is pre-constructed offline, and an elevator state mode library is generated through hierarchical association and integration, the method comprising:

[0029] The dynamic hierarchical sensitive indicator set is decomposed based on a hierarchical relationship to obtain a plurality of sensitive indicators; a specific indicator set of a first-level sensitive indicator is used as a screening condition to retrieve a plurality of sets of sample fault state time series data of a plurality of sample fault states at a plurality of sets of sample fault levels from the full-dimension operation log of the plurality of devices; a plurality of sets of state feature templates are output by performing indicator fluctuation scale analysis on the plurality of sets of sample fault state time series data; the plurality of sample fault states, the plurality of sets of sample fault levels, and the plurality of sets of sample state feature templates are hierarchically associated to obtain a first-level state mode sub-library; a plurality of levels of state mode sub-libraries are iteratively constructed in a hierarchical order, and a bidirectional mapping relationship index table of the plurality of levels of sensitive indicators and the plurality of levels of state mode sub-libraries is constructed to generate the elevator state mode library.

[0030] The dynamic hierarchical sensitive indicator set is decomposed based on a hierarchical relationship to obtain a plurality of sensitive indicators; a specific indicator set of a first-level sensitive indicator is used as a screening condition to retrieve a plurality of sets of sample fault state time series data of a plurality of sample fault states at a plurality of sets of sample fault levels from the full-dimension operation log of the plurality of devices; a plurality of sets of state feature templates are output by performing indicator fluctuation scale analysis on the plurality of sets of sample fault state time series data; the plurality of sample fault states, the plurality of sets of sample fault levels, and the plurality of sets of sample state feature templates are hierarchically associated to obtain a first-level state mode sub-library; a plurality of levels of state mode sub-libraries are iteratively constructed in a hierarchical order, and a bidirectional mapping relationship index table of the plurality of levels of sensitive indicators and the plurality of levels of state mode sub-libraries is constructed to generate the elevator state mode library.

[0031] According to the dynamic hierarchical sensitive indicator set, a plurality of levels of real-time operation parameters are retrieved from the power distribution cabinet in a priority order of sensitivity.

[0032] The sensitivity levels of all monitoring indicators in the dynamic hierarchical sensitive indicator set are obtained, and are ranked according to the sensitivity from high to low to form a sensitivity priority list. Subsequently, from the real-time data acquisition system of the power distribution cabinet, the real-time operating parameters of the corresponding indicators are retrieved in turn according to the sensitivity priority list, and the real-time data of the high-sensitivity indicators are preferentially collected as the first-level operating parameters; then the real-time data of the medium-sensitivity indicators are retrieved as the second-level operating parameters; and so on, the real-time operating parameters corresponding to each sensitivity level are retrieved in turn to form multi-level real-time operating parameters.

[0033] The multi-level real-time operating parameters are dynamically loaded into the elevator state mode library to trigger hierarchical equipment state mapping and output hierarchical state confidence vectors.

[0034] Further, the multi-level real-time operating parameters are dynamically loaded into the elevator state mode library to trigger hierarchical equipment state mapping and output hierarchical state confidence vectors, the method comprising:

[0035] The first-level operating parameter sequence of the first-level sensitive indicator is retrieved from the real-time data acquisition system of the power distribution cabinet; feature extraction is performed on the first-level operating parameter sequence to obtain a first-level feature vector; the first-level feature vector is loaded into the first-level state mode sub-library of the elevator state mode library to match and output an initial state mapping set; if the initial state mapping set is empty, a monitoring cycle of the first-level sensitive indicator is performed until the initial state mapping set is a non-empty set, triggering hierarchical equipment state mapping.

[0036] The first-level operating parameter sequence corresponding to the first-level sensitive indicator is retrieved from the real-time data acquisition system of the power distribution cabinet, which contains the continuous sampling data of the sensitive indicator within a preset time window; feature extraction is performed on the first-level operating parameter sequence to extract a multi-dimensional feature vector containing time domain and frequency domain features, specific features including mean, variance, kurtosis, skewness, spectral energy distribution, etc., to form a first-level feature vector; the first-level feature vector is loaded into the first-level state mode sub-library in the elevator state mode library, and the similarity between the first-level feature vector and each state feature template in the first-level state mode sub-library is calculated to match and output an initial state mapping set; if the initial state mapping set is empty, indicating that the current feature vector cannot match any known state template, the system continues to perform real-time monitoring of the first-level sensitive indicator, repeatedly retrieves the operating parameter sequence and performs feature extraction and matching, until a non-empty initial state mapping set is obtained, triggering subsequent hierarchical equipment state mapping processing.

[0037] Further, the first-level feature vector is loaded into the first-level state mode sub-library of the elevator state mode library to match and output an initial state mapping set, the method comprising:

[0038] The Euclidean distance matrix between the first-level feature vector and the plurality of sets of sample state feature templates in the first-level state pattern sub-library is iteratively calculated to output a plurality of state similarities; a maximum value is sorted in descending order within each of the plurality of sets of state similarities, and a plurality of real-time fault levels are retrieved from the plurality of sets of sample fault levels, and a plurality of maximum similarities are taken as a plurality of real-time fault probabilities of the plurality of real-time fault levels; based on a preset similarity threshold, the plurality of real-time fault probabilities are iterated to filter P real-time fault levels and P real-time fault probabilities of P sample fault states from the plurality of real-time fault levels; the P sample fault states, the P real-time fault levels and the P real-time fault probabilities are stored in association, and the initial state mapping set is output; if the plurality of real-time fault probabilities are all less than the preset similarity threshold, the initial state mapping set is an empty set.

[0039] First, the Euclidean distance matrix between the first-level feature vector and the plurality of sets of sample state feature templates in the first-level state pattern sub-library is iteratively calculated to measure the similarity between the input feature and each sample template; then, the Euclidean distance is converted into a similarity index (for example, by an inverse or exponential function mapping), and the similarity values of each template are sorted in descending order within each group to filter out a plurality of sample states corresponding to a plurality of maximum similarity values; according to the historical fault level information corresponding to these sample states, a plurality of real-time fault levels are retrieved, and the corresponding maximum similarity values are taken as the real-time fault probabilities of these fault levels to form a plurality of corresponding relationships between real-time fault levels and real-time fault probabilities; then, based on a preset similarity threshold, the above real-time fault probabilities are iterated to filter out the first P sample fault states with a similarity higher than the threshold, and the corresponding P real-time fault levels and real-time fault probabilities; finally, the P sample fault states, the corresponding P real-time fault levels and real-time fault probabilities filtered out are stored in association as an initial state mapping set output; if all real-time fault probabilities are lower than the preset similarity threshold, it is determined that the initial state mapping set is an empty set, indicating that the current input feature cannot be effectively matched with any known state template.

[0040] Further, the multi-level real-time running parameters are dynamically loaded into the elevator state pattern library to trigger hierarchical device state mapping, and a hierarchical state confidence vector is output. The method further comprises:

[0041] If the initial state mapping set is a non-empty set, a second-level sensitive index is called from the dynamic hierarchical sensitive index set; according to the index composition of the second-level sensitive index, after the second-level operation parameter sequence is called back from the power distribution cabinet historical data, the second-level feature vector is obtained through feature extraction; the first-level feature vector and the second-level feature vector are fused and loaded into the second-level state mode sub-library of the elevator state mode library, and the first verification state mapping set is output by matching the state feature templates of the sub-library; the state credibility of the initial state mapping set is verified by using the first verification state mapping set, and the first state confidence is output; if the first state confidence is higher than the pre-set confidence threshold, the initial state mapping set and the first verification state mapping set are fused by weighted fusion to generate the hierarchical state confidence vector; if the first state confidence is lower than the pre-set confidence threshold, the next level sensitive index is iteratively called for verification until the confidence meets the standard, and the hierarchical state confidence vector is output by integrating the multi-level mapping results.

[0042] If the initial state mapping set is a non-empty set, a second-level sensitive index set is called from the dynamic hierarchical sensitive index set; according to the index composition of the second-level sensitive index, the corresponding second-level operation parameter sequence is called back from the power distribution cabinet historical data, and the second-level feature vector is obtained by feature extraction; the first-level feature vector and the second-level feature vector are fused and loaded into the second-level state mode sub-library of the elevator state mode library, and the first verification state mapping set is output by matching the state feature templates of the sub-library; the state credibility of the initial state mapping set is verified by using the first verification state mapping set, and the first state confidence is output; when the first state confidence is higher than the pre-set confidence threshold, the initial state mapping set and the first verification state mapping set are fused based on the weighted fusion algorithm to generate the final hierarchical state confidence vector; if the first state confidence is lower than the pre-set confidence threshold, the subsequent lower level sensitive index is iteratively called, and the above verification process is repeated until the confidence meets the pre-set standard, and the comprehensive hierarchical state confidence vector is output by integrating the mapping results of each level.

[0043] Further, the state credibility of the initial state mapping set is verified by using the first verification state mapping set, and the first state confidence is output, which includes:

[0044] The first verification state mapping set and the initial state mapping set are subjected to state conflict detection, and the first conflict state mapping set and the initial conflict mapping set are screened and output; the first JSD distribution similarity of the first conflict state mapping set and the initial conflict mapping set is calculated as the first state confidence.

[0045] The fault state types and fault levels in the first verification state mapping set and the initial state mapping set are compared one by one, and state conflicts existing in the two mapping sets are identified, that is, the first conflict state mapping set and the initial conflict mapping set are screened out, and the conflict refers to a state pair in which the fault state types in the two mapping sets are inconsistent or the fault levels have significant differences. Based on the screened first conflict state mapping set and the initial conflict mapping set, the Jensen-Shannon divergence (JSD) distribution similarity of the corresponding fault state probability distribution is calculated, and the specific calculation steps include smoothing the two probability distributions, calculating the intermediate mixed distribution, and then obtaining the similarity value according to the definition of JSD. Finally, the JSD distribution similarity is taken as a measurement index of the first state confidence, reflecting the consistency and credibility between the initial state mapping set and the first verification state mapping set.

[0046] The hierarchical state confidence vector is subjected to state transition prediction, and a state transition probability matrix is output.

[0047] Further, the hierarchical state confidence vector is subjected to state transition prediction, and a state transition probability matrix is output, and the method comprises:

[0048] A plurality of sample state confidence vector sequences are obtained interactively, and a state transition frequency chain is output by counting the state transition frequencies of the plurality of sample state confidence vector sequences. A reference state transition frequency matrix is constructed based on the state transition frequency chain. The state transition probability of the hierarchical state confidence vector is normalized and corrected based on the reference state transition frequency matrix, and the state transition probability matrix is output.

[0049] State confidence vector sequences of a plurality of historical samples are obtained interactively, and the state confidence vector sequences cover the time evolution process of normal and multiple fault states. The transition relationship between adjacent states in the plurality of sample state confidence vector sequences is counted to obtain a state transition frequency chain, which records the transition times and frequencies between different states. Based on the state transition frequency chain, a reference state transition frequency matrix is constructed, and the elements in the matrix represent the transition frequencies between states. The rows of the matrix represent the current state, and the columns represent the possible next state. For the hierarchical state confidence vector at the current time, the state transition probability is normalized and corrected based on the reference state transition frequency matrix, and the state transition probability matrix is generated after correction, reflecting the transition probability distribution between states. Finally, the state transition probability matrix is output, providing a basis for subsequent real-time fault probability distribution calculation and multi-level early warning signal generation.

[0050] The hierarchical state confidence vector and the state transition probability matrix are fused, and a real-time fault probability distribution is output.

[0051] Firstly, the hierarchical state confidence vector of the current moment is taken as the initial probability distribution of the fault state; then, the initial probability distribution is weighted and corrected and time series predicted according to the state transition probability matrix, the conditional transition probability of each fault state is calculated, and the state probability distribution adjusted based on the historical transition rule is obtained; subsequently, the corrected probability distribution and the initial confidence vector are fused, and a comprehensive real-time fault probability distribution is generated through weighted average or Bayesian update and the like; finally, the real-time fault probability distribution is output to support fault diagnosis, risk assessment and generation of multi-level warning signals.

[0052] According to the real-time fault probability distribution, a multi-level warning signal is generated to trigger a gradient safety protection response.

[0053] Based on the probability values corresponding to each fault state in the real-time fault probability distribution, multi-level warning thresholds are set, respectively corresponding to low-level warning, medium-level warning and high-level warning levels; the probability of each fault state is compared with the preset threshold to determine the current fault risk level; when the fault probability exceeds the corresponding threshold, a warning signal of the corresponding level is generated; according to the generated multi-level warning signal, the safety protection response measures of the corresponding gradient are triggered, including but not limited to alarm prompt, operation limitation, automatic shutdown and emergency fault handling program start and the like.

[0054] In summary, the embodiments of the present application have at least the following technical effects:

[0055] Firstly, the dynamic hierarchical sensitive index set is obtained by verifying the state sensitivity of the multiple monitoring indexes of the power distribution cabinet. Then, the multi-level state mode sub-library is pre-constructed offline according to the dynamic hierarchical sensitive index set, and the elevator state mode library is generated through hierarchical association integration; at the same time, the multi-level real-time operation parameters are retrieved from the power distribution cabinet in order of sensitivity priority according to the dynamic hierarchical sensitive index set. Further, the multi-level real-time operation parameters are dynamically loaded into the elevator state mode library to trigger hierarchical device state mapping and output the hierarchical state confidence vector. Next, the hierarchical state confidence vector is subjected to state transition prediction to output the state transition probability matrix. Then, the hierarchical state confidence vector and the state transition probability matrix are fused to output the real-time fault probability distribution. Finally, the multi-level warning signal is generated according to the real-time fault probability distribution to trigger the gradient safety protection response. The technical problem of relying on sensors for state recognition of special equipment in the prior art is solved, the accurate recognition of the running state of the elevator based on electrical parameters is realized, thereby reducing the dependence on external sensors and improving the real-time performance and reliability of state monitoring.

[0056] Embodiment two, based on the same inventive concept as the special equipment state recognition method based on electrical parameter pattern matching in the foregoing embodiments, such as Figure 2As shown, the application provides a special equipment state recognition system based on electrical parameter pattern matching, wherein the system comprises:

[0057] The sensitivity verification module 11 verifies the sensitivity of the multiple monitoring indicators of the power distribution cabinet to the state of the special equipment, and obtains a dynamic hierarchical sensitive indicator set; the pattern library construction module 12 pre-constructs multiple-level state pattern sub-libraries offline according to the dynamic hierarchical sensitive indicator set, and integrates them through hierarchical association to generate an elevator state pattern library; the parameter calling module 13 calls multiple-level real-time operating parameters from the power distribution cabinet in order of sensitivity priority according to the dynamic hierarchical sensitive indicator set; the state mapping module 14 dynamically loads the multiple-level real-time operating parameters into the elevator state pattern library, triggers hierarchical equipment state mapping, and outputs a hierarchical state confidence vector; the state transition prediction module 15 performs state transition prediction on the hierarchical state confidence vector, and outputs a state transition probability matrix; the fault probability output module 16 fuses the hierarchical state confidence vector and the state transition probability matrix, and outputs a real-time fault probability distribution; the early warning response module 17 generates a multi-level early warning signal according to the real-time fault probability distribution, and triggers a ladder safety protection response.

[0058] Further, the sensitivity verification module 11 is configured to perform the following method:

[0059] The application environment features of the power distribution cabinet are called, and the application environment features and the model code of the power distribution cabinet are used as device fingerprint coding retrieval conditions to call multiple device full-dimensional operating logs; the multiple device full-dimensional operating logs are aggregated based on the index type to obtain multiple multi-state time series data sets of multiple operating indicators; the sensitivity of the special equipment state is quantified based on the multiple multi-state time series data sets, and multiple operating comprehensive sensitivities are output; the multiple operating indicators are dynamically classified according to the multiple operating comprehensive sensitivities, and the dynamic hierarchical sensitive indicator set is output.

[0060] Further, the pattern library construction module 12 is configured to perform the following method:

[0061] The dynamic hierarchical sensitive indicator set is decomposed based on the hierarchical relationship to obtain multiple-level sensitive indicators; a specific indicator set of a first-level sensitive indicator is used as a screening condition to call multiple sets of sample fault state time series data sets of multiple sample fault states at multiple sets of sample fault levels from the multiple device full-dimensional operating logs; multiple sets of state feature templates are output by performing index fluctuation scale analysis on the multiple sets of sample fault state time series data sets; the multiple sample fault states, the multiple sets of sample fault levels, and the multiple sets of sample state feature templates are stored in a hierarchical association to obtain a first-level state pattern sub-library; multiple-level state pattern sub-libraries are iteratively constructed in a hierarchical order, and a bidirectional mapping relationship index table of the multiple-level sensitive indicators and the multiple-level state pattern sub-libraries is constructed to generate the elevator state pattern library.

[0062] Further, the state mapping module 14 is configured to execute the following method:

[0063] The first-level operating parameter sequence of the first-level sensitive index is retrieved from the power distribution cabinet in real time, the first-level operating parameter sequence is subjected to feature extraction to obtain a first-level feature vector, the first-level feature vector is loaded into a first-level state mode sub-library of the elevator state mode library, and an initial state mapping set is outputted by matching; if the initial state mapping set is an empty set, a monitoring cycle of the first-level sensitive index is performed until the initial state mapping set is a non-empty set, triggering hierarchical equipment state mapping.

[0064] Further, the state mapping module 14 is configured to execute the following method:

[0065] If the initial state mapping set is a non-empty set, a second-level sensitive index is retrieved from the dynamic hierarchical sensitive index set, a second-level operating parameter sequence is retrieved from the power distribution cabinet according to the index composition of the second-level sensitive index, a second-level feature vector is obtained by feature extraction, the first-level feature vector and the second-level feature vector are fused and loaded into a second-level state mode sub-library of the elevator state mode library, and a first verification state mapping set is outputted by matching; the state credibility of the initial state mapping set is verified by using the first verification state mapping set, and a first state confidence is outputted; if the first state confidence is higher than a preset confidence threshold, the initial state mapping set and the first verification state mapping set are fused by weighting to generate the hierarchical state confidence vector; if the first state confidence is lower than the preset confidence threshold, a next-level sensitive index is iteratively called for verification until the confidence meets the standard, and the hierarchical state confidence vector is outputted by integrating the mapping results of multiple levels.

[0066] Further, the state mapping module 14 is configured to execute the following method:

[0067] The Euclidean distance matrix of the first-level feature vector and multiple groups of sample state feature templates in the first-level state mode sub-library is calculated, and multiple groups of state similarities are outputted; according to the descending maximum value of the multiple groups of state similarities, multiple real-time fault levels are retrieved from the multiple groups of sample fault levels, and multiple maximum similarities are taken as multiple real-time fault probabilities of the multiple real-time fault levels; based on a preset similarity threshold, the multiple real-time fault probabilities are traversed, P real-time fault levels and P real-time fault probabilities of P sample fault states are screened from the multiple real-time fault levels, the P sample fault states, the P real-time fault levels and the P real-time fault probabilities are stored in association, and the initial state mapping set is outputted; if the multiple real-time fault probabilities are all less than the preset similarity threshold, the initial state mapping set is an empty set.

[0068] Further, the state mapping module 14 is configured to perform the following method:

[0069] Performing state conflict detection on the first verification state mapping set and the initial state mapping set, and screening to output a first conflict state mapping set and an initial conflict mapping set; calculating the first JSD distribution similarity of the first conflict state mapping set and the initial conflict mapping set as the first state confidence.

[0070] Further, the state transition prediction module 15 is configured to perform the following method:

[0071] Obtaining a plurality of sample state confidence vector sequences through interaction, and performing state transition frequency statistics on the plurality of sample state confidence vector sequences to output a state transition frequency chain; constructing a benchmark state transition frequency matrix based on the state transition frequency chain; and performing state transition probability normalization correction on the hierarchical state confidence vector based on the benchmark state transition frequency matrix to output the state transition probability matrix.

[0072] Further, the sensitivity verification module 11 is configured to perform the following method:

[0073] Based on the device state label, decompose the plurality of multi-state time series data sets to obtain a plurality of benchmark state record sets and a plurality of groups of fault state record sets; based on the plurality of benchmark state record sets, quantize the plurality of groups of fault state record sets using KL divergence to output a plurality of groups of fault state KL sensitivity scores; and according to the recurrence frequency of M fault states, perform group weighting on the plurality of groups of fault state KL sensitivity scores to output the plurality of operation comprehensive sensitivities.

[0074] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes a specific embodiment of the present application. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or may be advantageous.

[0075] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0076] The present specification and drawings are only exemplary descriptions of the present application, and are considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalents, the present application intends to include these modifications and changes.

Claims

1. A method for identifying the state of special equipment based on matching of electrical parameter patterns, characterized in that, The method includes: Special equipment status sensitivity verification was performed on multiple monitoring indicators of the power distribution cabinet to obtain a dynamic hierarchical sensitive indicator set; Based on the aforementioned dynamic hierarchical sensitive index set, a multi-level state pattern sub-library is pre-constructed offline, and an elevator state pattern library is generated through hierarchical association integration. Based on the dynamic hierarchical sensitive index set, multi-level real-time operating parameters are retrieved from the power distribution cabinet according to sensitivity priority. The multi-level real-time operating parameters are dynamically loaded into the elevator state mode library, triggering hierarchical equipment state mapping and outputting a hierarchical state confidence vector. Perform state transition prediction on the hierarchical state confidence vector and output the state transition probability matrix; By fusing the hierarchical state confidence vector and the state transition probability matrix, a real-time fault probability distribution is output. Based on the real-time fault probability distribution, multi-level early warning signals are generated to trigger tiered safety protection responses. Special equipment status sensitivity verification was performed on multiple monitoring indicators of the distribution cabinet to obtain a dynamic hierarchical sensitivity indicator set, including: The application environment characteristics of the power distribution cabinet are retrieved, and the application environment characteristics and the power distribution cabinet model code are used as the equipment fingerprint code retrieval conditions to call up the full-dimensional operation logs of multiple devices. Based on the index type, the full-dimensional operation logs of the multiple devices are aggregated to obtain multiple multi-state time series datasets of multiple operation indicators; Based on the multiple multi-state time series datasets, special equipment state sensitivity quantification is performed, and multiple comprehensive operational sensitivity values ​​are output. Based on the multiple comprehensive operational sensitivities, the multiple operational indicators are dynamically classified, and the dynamically classified sensitivity indicator set is output. Based on the aforementioned dynamic hierarchical sensitive index set, a multi-level state pattern sub-library is pre-constructed offline, and an elevator state pattern library is generated through hierarchical association integration, including: Based on the hierarchical relationship, the dynamic hierarchical sensitive index set is decomposed to obtain multi-level sensitive indicators; Using the specific set of first-level sensitive indicators as the filtering criteria, retrieve time-series datasets of multiple sample fault states under multiple sample fault levels from the full-dimensional operation logs of the multiple devices. By performing index fluctuation scale analysis on the time series dataset of the multiple sets of sample fault states, multiple sets of state feature templates are output. The hierarchical association storage of the various sample fault states, multiple sets of sample fault levels, and multiple sets of sample state feature templates yields the first-level state pattern sub-library. A multi-level state pattern sub-library is iteratively constructed in hierarchical order, and a bidirectional mapping index table between the multi-level sensitive indicators and the multi-level state pattern sub-library is constructed to generate the elevator state pattern library.

2. The special equipment state recognition method based on electrical parameter pattern matching according to claim 1, characterized in that, The method involves dynamically loading the multi-level real-time operating parameters into the elevator state pattern library, triggering hierarchical equipment state mapping, and outputting a hierarchical state confidence vector. The first-level operating parameter sequence of the first-level sensitive indicator is retrieved in real time from the power distribution cabinet; Feature extraction is performed on the first-level operating parameter sequence to obtain the first-level feature vector; The first-level feature vector is loaded into the first-level state pattern sub-library of the elevator state pattern library, and the initial state mapping set is matched and output. If the initial state mapping set is empty, a monitoring cycle of the first level sensitive indicator is performed until the initial state mapping set is non-empty, triggering hierarchical equipment state mapping.

3. The special equipment state recognition method based on electrical parameter pattern matching according to claim 2, characterized in that, The multi-level real-time operation parameter is dynamically loaded to the elevator state mode library, triggering hierarchical equipment state mapping, and outputting a hierarchical state confidence vector. The method further comprises: If the initial state mapping set is non-empty, a second level sensitive indicator is called from the dynamic hierarchical sensitive indicator set; According to the indicator composition of the second level sensitive indicator, after the power distribution cabinet backtracks to call a second level operation parameter sequence, a second level feature vector is obtained through feature extraction; The first level feature vector and the second level feature vector are fused and loaded to a second level state mode sub-library of the elevator state mode library, and a first verification state mapping set is matched and output; The first verification state mapping set is used to verify the state credibility of the initial state mapping set, and a first state confidence is output; If the first state confidence is higher than a preset confidence threshold, the initial state mapping set and the first verification state mapping set are weighted and fused to generate the hierarchical state confidence vector; If the first state confidence is lower than the preset confidence threshold, a next level sensitive indicator is iteratively called for verification until the confidence meets the standard, and a hierarchical state confidence vector is output by integrating the multi-level mapping results.

4. The special equipment state recognition method based on electrical parameter pattern matching according to claim 2, characterized in that, The first level feature vector is loaded to a first level state mode sub-library of the elevator state mode library, and an initial state mapping set is matched and output. The method comprises: Euclidean distance matrices of multiple groups of sample state feature templates in the first level feature vector and the first level state mode sub-library are calculated, and multiple groups of state similarities are output; According to the descending maximum values of the multiple groups of state similarities, multiple real-time fault levels are called from the multiple groups of sample fault levels, and multiple maximum similarities are taken as multiple real-time fault probabilities of the multiple real-time fault levels; Based on a preset similarity threshold, the multiple real-time fault probabilities are traversed, P real-time fault levels and P real-time fault probabilities of P sample fault states are selected from the multiple real-time fault levels; The P sample fault states, the P real-time fault levels and the P real-time fault probabilities are stored in association, and the initial state mapping set is output; If the multiple real-time fault probabilities are all less than the preset similarity threshold, the initial state mapping set is empty.

5. The special equipment state recognition method based on electrical parameter pattern matching of claim 3, wherein, The first verification state mapping set is used to verify the state credibility of the initial state mapping set, and a first state confidence is output. The method comprises: State conflict detection is performed on the first verification state mapping set and the initial state mapping set, and a first conflict state mapping set and an initial conflict mapping set are selected and output; A first JSD distribution similarity of the first conflict state mapping set and the initial conflict mapping set is calculated as the first state confidence.

6. The special equipment state recognition method based on electrical parameter pattern matching of claim 1, wherein, State transition prediction is performed on the hierarchical state confidence vector, and a state transition probability matrix is output. The method comprises: Interactively obtain a plurality of sample state confidence vector sequences, and output a state transition frequency chain by performing state transition frequency statistics on the plurality of sample state confidence vector sequences; Construct a benchmark state transition frequency matrix based on the state transition frequency chain; Perform state transition probability normalization correction of the hierarchical state confidence vector based on the benchmark state transition frequency matrix, and output the state transition probability matrix.

7. The special equipment state recognition method based on electrical parameter pattern matching of claim 1, wherein, Quantify the state sensitivity of special equipment based on the plurality of multi-state time series data sets, and output a plurality of operation comprehensive sensitivities, the method comprising: Decompose the plurality of multi-state time series data sets based on equipment state labels to obtain a plurality of benchmark state record sets and a plurality of sets of fault state record sets; Quantify the plurality of sets of fault state record sets using KL divergence based on the plurality of benchmark state record sets, and output a plurality of sets of fault state KL sensitivity scores; According to the recurrence frequency of M fault states, intra-group weighting is performed on the plurality of sets of fault state KL sensitivity scores, and the plurality of operation comprehensive sensitivities are output.

8. A special equipment state recognition system based on electrical parameter pattern matching, characterized in that, A system for implementing the special equipment state recognition method based on electrical parameter pattern matching of any one of claims 1-7, the system comprising: A sensitivity verification module: performing special equipment state sensitivity verification on multiple monitoring indicators of a power distribution cabinet to obtain a dynamic hierarchical sensitive index set; A pattern library construction module: constructing a multi-level state pattern sub-library offline according to the dynamic hierarchical sensitive index set, and integrating to generate an elevator state pattern library through hierarchical association; A parameter retrieval module: retrieving multi-level real-time operation parameters from the power distribution cabinet according to the dynamic hierarchical sensitive index set in order of sensitivity priority; A state mapping module: dynamically loading the multi-level real-time operation parameters into the elevator state pattern library to trigger hierarchical equipment state mapping and output a hierarchical state confidence vector; A state transition prediction module: performing state transition prediction on the hierarchical state confidence vector to output a state transition probability matrix; A fault probability output module: fusing the hierarchical state confidence vector and the state transition probability matrix to output a real-time fault probability distribution; An early warning response module: generating a multi-level early warning signal according to the real-time fault probability distribution to trigger a hierarchical safety protection response.

Images